# Examples that must not share the machine with other examples, one
# "script.py: reason" per line. A reason is mandatory and an entry naming a
# deleted script fails the meta-test, so this list stays honest.
#
# These are NOT skipped. The example gate runs its bulk under pytest-xdist and
# then gives everything listed here a second, SERIAL pass -- the same split,
# for the same reason, that `make test-python` already applies to the
# benchmark directories. An entry here costs wall-clock, never coverage.
#
# The bar for an entry is narrow: the example's own assertion must be a
# TIMING. An example that is merely slow does not belong here (it parallelizes
# fine); an example whose measurement is perturbed by load does.
ddc_fn_scaling.py: asserts a thread-scaling speedup to prove execute() releases the GIL -- under xdist the workers already own every core, so the whole curve is depressed and it read the contention as "execute appears GIL-bound" (measured 1.15x at 2 threads). Reading the best point rather than the 2-thread one (doppler#990) fixed a DIFFERENT failure and does not remove this one: xdist starves every count at once, so there is no unaffected point to find
